Informant-based hearing difficulties and the risk for mild cognitive impairment and dementia

Abstract

Background: hearing loss has been associated with mild cognitive impairment (MCI) and dementia. Studies have not assessed whether hearing difficulties (HD) that interfere with daily activities as reported by partners can be a marker for increased risk for cognitive decline and impairment.

Objective: to assess the cross-sectional and longitudinal associations between informant-based HD, which interfere with daily activities and the risk for MCI and dementia.

Methods: the study included 4812 participants without dementia, enrolled in the Mayo Clinic Study of Aging (mean age (SD) 73.7 (9.6) years) with cognitive evaluation and informant-based report on participant’s HD that interfere significantly with daily activities at baseline and for every 15 months. Cox proportional hazards models (utilising time-dependent HD status and age as the time scale) were used to examine HD and the risk for MCI or dementia, and mixed-effects models (allowing for random subject-specific intercepts and slopes) were used to examine the relationship between HD and cognitive decline.

Results: about, 981 participants had HD and 612 (12.7%) had prevalent MCI at baseline; 759 participants developed incident MCI and 273 developed incident dementia. In cognitively unimpaired participants at baseline, those with HD had higher risk for MCI (hazard ratio [HR] = 1.29, 95% confidence interval [CI] (1.10, 1.51), P = 0.002; adjusting for sex, years of education). In participants without dementia, those with HD had higher risk for dementia (HR: 1.39, 95% CI, (1.08–1.79), P = 0.011; adjusting sex and education). In individuals with MCI, HD was associated with modestly greater cognitive decline.

Conclusions: informant-based HD was associated with increased risk for MCI and dementia.

Key points

• Individuals with informant-based significant HD had increased risk for MCI and dementia.

• Participants with MCI and HD had modestly greater cognitive decline than those with MCI but without HD.

• Informant-based HD information is easily obtained and could contribute to lead to interventions earlier.

Introduction

Age-related impairments pose significant barriers to an individual’s quality of life, social interactions, independence or personal productivity [1]. Aging is associated with cognitive and sensory decline [2]. Hearing difficulty is a common social and medical problem with 10% of the population experiencing hearing loss that impairs communication; two-thirds of individuals older than 70 years have clinically meaningful hearing impairment that could affect daily communication [3]. As individuals age, they also typically experience declines in their cognitive abilities [4], which are heterogeneous across individuals [5], and range from typical aging to mild cognitive impairment (MCI) to Alzheimer’s disease (AD) and other dementias [6]. Previous reports suggest that greater hearing loss severity is associated with poorer cognitive function [7–9], poorer cognitive performance on auditory and non-auditory tests [10–12], faster cognitive decline [7, 12] and incident dementia [12–15].

For most older adults, change in hearing is gradual (e.g. pure-tone thresholds increase approximately 1 dB every year) and for someone to detect a change in their own hearing would require them to have a sharp cognitive function. In the context of an aging person, the concern of hearing difficulty is either reported by the patients themselves or by the partners (typically, a spouse or friends). Self-reported hearing loss is associated with cognitive decline in older persons [16], but studies have not assessed whether hearing difficulties (HD) as reported by partners can be a marker for increased risk for cognitive decline and impairment. In the current study, we aimed to examine the cross-sectional and longitudinal associations of informant-based HD with MCI and dementia.

Methods

Study population

This is a prospective cohort study of participants without dementia (at baseline), evaluated in person by the Mayo Clinic Study of Aging (MCSA), with available information on HD and at least a cognitive evaluation at baseline (N = 4812). The study design and methodology of the MCSA have previously been published in detail [17, 18]. In summary, Olmsted County (MN) residents, 70–89 years of age on 1 October 2004, were enumerated using the resources of the Rochester Epidemiology Project (REP) medical records linkage system [19] and an age and sex-stratified random sample of eligible residents (without dementia, not terminally ill or in hospice) was selected and invited to participate in-person or by telephone. Ongoing recruitment was initiated in 2008, and in 2012, MCSA began recruitment of participants at age 50–69 years. Follow-up was performed at 15-month intervals using the same clinical protocol for evaluation and diagnosis as at baseline. The study was approved by the Institutional Review Boards of the Mayo Clinic and of the Olmsted Medical Center. Written informed consent was obtained prior to participation in the study.

Clinical evaluation

MCSA participants were evaluated by a study coordinator (SC), a physician and underwent neuropsychological testing by a trained psychometrist. Demographic information was ascertained by interview, and questions about memory were administered to the participant by the SC. In addition, the SC administered questions and questionnaires to an informant (a study partner, usually spouse or another individual that knew the participant very well) related to the participant’s health and cognitive function. The physician evaluation included a medical history review, a neurologic examination and administration of the short test of mental status (STMS) [20]. Neuropsychological testing included nine tests to assess performance in four cognitive domains: memory, attention/executive function, language and visuospatial skills [17, 18]. Individual test scores were converted to z-scores (using the mean and standard deviation [SD] from the MCSA 2004 enrolment cohort). The individual z-scores were then averaged to generate four cognitive domain scores which we also converted to z-scores. Finally, from the average of the four domain z-scores, we created a global cognitive score, which was converted to a z-score (by subtracting the mean and dividing by the SD). Cognitive impairment was considered if the domain score was ≥1.0 SD below the mean.

In a weekly meeting, the CS, the physician and a neuropsychologist reviewed all information for each participant, and a diagnosis of MCI [21], dementia [22] or normal cognition was made by consensus. If impairment occurred in the memory domain then the participant was classified as having amnestic MCI (aMCI) and if impairment occurred only in non-memory domains (1 or more) then the participant was classified as having non-amnestic MCI (naMCI). Participants were classified as cognitively unimpaired (CU) if they performed in the normative range and did not meet criteria for MCI or dementia. Evaluators were not aware of clinical and cognitive findings from previous evaluations when making a diagnosis during follow-up visits.

Identification of HD

Each MCSA participant was asked to suggest a study partner as an informant in the study. Usually, informants were spouses or other individuals that knew the participants well. HD interfering with daily activities were based on the informant’s perception as s/he answered the following question: ‘Does (the participant) have significant HD that interfere with daily activities (yes/no/do not know)’. This question was asked at baseline and each follow-up visit. In the present manuscript the abbreviation HD represents the exact question asked to the informant.

Statistical analysis

Demographics were compared between those with and without HD using Kruskal–Wallis tests or χ² tests as appropriate. For analytical assessment of performance, the raw scores for tests in each cognitive domain were z-scored, averaged, and scaled to create domain-specific cognitive z-scores. In addition, a global z-score for overall cognitive performance was also created by averaging and scaling the four domain z-scores. Hazard ratio (HR) for incident MCI or dementia obtained from Cox proportional hazards models utilising time-dependent HD status was adjusted for sex and years of education, with age as the time scale; we used the most recent recorded HD value in predicting later outcome. Adjusting for APOE ε4 carriership did not change the estimates appreciably and is not included in tables. Interaction by sex was examined. If a participant progressed to dementia without an MCI diagnosis, s/he was presumed to have passed through an undetected MCI phase and was included as incident MCI events, as well, in analysis having incident MCI. Separate models adjusting additionally for baseline global z-score or STMS were also built to examine the associations while adjusting for the participant’s cognitive performance. Mixed-effects models, adjusted for age, sex, and education, were used to examine the association between baseline HD (yes/no) and change in z-scored global cognition, allowing for random subject-specific intercepts and slopes. All analyses were considered statistically significant at a P-value ≤0.05 and were performed using the SAS statistical software version 9.4 (SAS Institute, Cary, North Carolina).

Results

Participants’ characteristics

Among 4812 participants (Table 1) without dementia at baseline [mean age (SD): 73.7 (9.6); 48.5% females], 981 had HD (20.4%), 1240 (27.2%) were APOE ε4 positive and 612 (12.7%) had MCI. During the follow-up [mean (SD): 5.4 (3.2) years], 273 (6.8%) developed incident dementia among participants without dementia at baseline who had at least one follow-up evaluation (N = 4011) and 759 developed incident MCI [follow-up mean (SD): 5.1 (3.2) years] among the CU at baseline who had at least one follow-up evaluation (N = 3536). Most of the informants (64.1%) were spouses, followed by children (20.4%), friend/companion (9.0%), other relatives (4.5%), son-/daughter-in-law (0.9%) and paid caregiver (0.2%). About, 68% of informants reported living with the participant (N = 3273) and 73% of informants reported seeing the participant daily (N = 3511).

Table 1.

Participants’ characteristics by significant HD at baseline

	Significant HD at baseline
	No (N = 3831)	Yes (N = 981)	Total (N = 4812)	P-valuea
Age, at baseline	73.0 (9.7)	76.3 (8.7)	73.7 (9.6)	<0.001
Sex, female	2059 (53.7)	274 (27.9)	2333 (48.5)	<0.001
Education, years	14.3 (2.8)	13.9 (2.9)	14.3 (2.8)	<0.001
APOE ε4 positiveb	990 (27.2)	250 (27.0)	1240 (27.2)	0.87
Global cognitive z-scorec	−0.3 (1.2)	−0.8 (1.2)	−0.4 (1.2)	<0.001
Memory z-scorec	−0.3 (1.1)	−0.7 (1.1)	−0.4 (1.1)	<0.001
Language z-scorec	−0.3 (1.2)	−0.6 (1.2)	−0.3 (1.2)	<0.001
Attention/executive z-scorec	−0.3 (1.2)	−0.8 (1.3)	−0.4 (1.2)	<0.001
Visuospatial skills z-scorec	−0.2 (1.1)	−0.4 (1.1)	−0.3 (1.1)	<0.001
Cognitive status				<0.001
CU	3411 (89.0)	789 (80.4)	4200 (87.3)
MCI	420 (11.0)	192 (19.6)	612 (12.7)
Type of MCI, at baseline				0.42
Amnestic	311 (74.0)	148 (77.1)	459 (75.0)
Non-amnestic	109 (26.0)	44 (22.9)	153 (25.0)
Incident MCId	583 (20.2)	176 (27.2)	759 (21.5)	<0.001
Years to MCI or last follow-up	5.1 (3.2)	4.9 (3.3)	5.1(3.2)	0.12
No of visits to MCI or last follow-up	4.9 (2.4)	4.9 (2.5)	4.9 (2.4)	0.74
Incident dementiae	182 (5.6)	91 (11.5)	273 (6.8)	<0.001
Years to dementia or last follow-up	5.5 (3.2)	5.2 (3.3)	5.4 (3.2)	0.02
No of visits to dementia or last follow-up	5.1 (2.4)	5.0 (2.5)	5.1 (2.4)	0.33
Type of informantf				<0.001
Spouse	2382 (62.2)	700 (71.4)	3082 (64.1)
Otherg	1448 (37.8)	281 (28.6)	1729 (35.9)

Note: data presented as N (%) for categorical and mean (SD) for continuous characteristics.

^at-test or χ² tests as appropriate.

^b251 missing (54 with HD).

^cGlobal cognitive z-score was computed after scaling raw cognitive test scores (mean 0 ± 1) using data for CU participants at baseline. Domain-specific z-scores were summed and scaled to obtain global z-scores.

^dCases of dementia going from CU straight to dementia are included; 3536 were CU at baseline with follow-up for MCI outcome (647 participants with HD).

^eCases of dementia going from CU straight to dementia are included; 4011 were non-demented at baseline with follow-up for dementia outcome (788 participants with HD).

^fOne missing (zero participants with HD).

^gChildren (20.4%), in-law son/daughter (0.9%), other relatives (4.5%), friend/companion (9.0%) and paid caregiver (0.2%).

HD, MCI and dementia

In cross-sectional analysis at baseline, participants with HD had higher odds of MCI compared with participants without HD [odds ratio (OR) = 1.50, 95% CI (1.23, 1.83), P < 0.001]. The OR estimate was stronger in women [OR = 1.78, 95% CI (1.27, 2.49), P < 0.001] than men [OR = 1.37, 95% CI (1.07, 1.75), P = 0.012], although the interaction by sex was not statistically significant (P = 0.28).

In CU participants at baseline, those with HD had a higher risk for incident MCI (HR = 1.29, 95% CI (1.10, 1.51), P = 0.002, adjusting for sex and education; Table 2). Participants without dementia at baseline (i.e. including CU and those with MCI) who had HD had also a higher risk for incident dementia [vs. participants without dementia and without HD; HR: 1.39, 95% CI (1.08–1.79), P = 0.011]. This estimate was stronger for those who were CU at baseline [HR: 1.72, 95% CI (1.20–1.27), P = 0.003], but was not increased for participants with prevalent (at baseline) or incident (during follow-up) MCI. Adjustment for cognitive performance at baseline (measured by global z-score (for overall cognitive performance) or STMS) minimally changed estimates (Table 2).

Table 2.

Association between hearing difficulties and incident MCI or dementia

Outcome	Events	HR (95% CI)a	P-value
CU to MCIb	759	1.29 (1.10, 1.51)	0.002
Additionally adjusted for global z-scorec		1.34 (1.14, 1.58)	<0.001
Additionally adjusted for STMS		1.23 (1.05, 1.44)	0.011
CU to aMCId,e	484	1.23 (1.01, 1.50)	0.040
Additionally adjusted for global z-score		1.29 (1.05, 1.59)	0.014
Additionally adjusted for STMS		1.17 (0.96, 1.43)	0.12
CU to naMCId,e	175	1.39 (1.01, 1.92)	0.046
Additionally adjusted for global z-score		1.51 (1.08, 2.11)	0.017
Additionally adjusted for STMS		1.33 (0.96, 1.86)	0.09
ND to dementiae	275	1.39 (1.08, 1.79)	0.011
Additionally adjusted for global z-score		1.33 (1.02, 1.74)	0.038
Additionally adjusted for STMS		1.30 (1.01, 1.68)	0.040
CU to dementiab	130	1.72 (1.20, 2.47)	0.003
Additionally adjusted for global z-score		1.67 (1.14, 2.45)	0.009
Additionally adjusted for STMS		1.72 (1.20, 2.47)	0.003
All MCI to dementiaa	241	0.95 (0.72, 1.24)	0.68
Incident MCI to dementia		0.97 (0.63, 1.48)	0.87
Prevalent MCI to dementia		1.02 (0.72, 1.46)	0.90

Note: CU stands for cognitively unimpaired; MCI for mild cognitive impairment; aMCI for amnestic MCI; naMCI for non-amnestic MCI and STMS for short test of mental status.

^aHR (95% CI) retained from Cox proportional hazards models utilising time-dependent hearing loss status adjusted for sex and years of education, with age as the time scale.

^bIn CU at baseline; if a participant progressed to dementia without an MCI diagnosis at an MCSA evaluation, it was presumed he/she has passed through an undetected MCI phase and was included as incident MCI events, as well, in analysis having as outcome incident MCI.

^cThe raw scores for tests in each cognitive domain were z-scored, averaged and scaled to create domain-specific cognitive z-scores. A global z-score for overall cognitive performance was also created by averaging and scaling the four domain z-scores.

^dParticipants who progressed to dementia without an MCI diagnosis at an MCSA evaluation were not included, as they could not be characterised as aMCI or naMCI.

^eIncluding all non-demented participants at baseline.

HD and cognitive decline

During the follow-up period, in the total population without dementia, there was a statistically significant difference in the annualised change in the language cognitive z-score (with vs. without HD at baseline; annualised change ± SE: −0.012 ± 0.007, P = 0.05; Appendix Table S1, available as supplementary data in Age and Ageing online) and the attention/executive function cognitive z-score (with vs. without HD at baseline; annualised change ± SE: −0.014 ± 0.006, P = 0.03). We also examined the change over time in cognitive z-scores by cognitive group at baseline (i.e. CU, MCI). CU individuals with HD at baseline declined similarly with regards to global cognition z-scores compared to CU without HD (estimated annualised difference −0.001; 95% CI, −0.007 to 0.005; P = 0.86). In contrast, slopes in participants with MCI and HD at baseline were steeper than in those with MCI but without HD (estimated annualised difference −0.047; 95% CI, −0.064 to −0.030; P = 0.006; Figure 1). However, overall, the annualised changes associated with HD were modest.

Figure 1.

Longitudinal association of HD and cognitive performance. In mixed-effects models adjusting for age, sex, education and allowing for random subject-specific intercepts and slopes, CU individuals with baseline HD declined similarly with regards to global cognition z-scores compared to CU without HD (estimated annualised difference −0.001; 95% CI, −0.007 to 0.005; P = 0.86), while slopes in participants with MCI and HD were steeper than in those with MCI but without HD (estimated annualised difference −0.047; 95% CI, −0.064 to −0.030; P = 0.006). Figure derived from mixed-effects models using mean covariate values for age, sex and education, allowing for random subject-specific intercepts and slopes.

Discussion

Informant-based HD was associated with increased risk for MCI and dementia in those without dementia and CU individuals, and modestly greater cognitive decline during follow-up in participants with MCI. The current study utilising partner-based HD information is consistent with prior studies examining the relationship of hearing loss (evaluated by other means) and cognitive impairment [7–16].

Family and acquaintances provide strong motivation to seek help for HD interventions; [23]; however, audiometric thresholds do not always reflect the communication difficulties individuals have [24] as the ability of daily communication depends on more than sound audibility. We hypothesised that informant-based HD would be another useful and easily obtained source of information because it is not based on a single assessment or observation but accounts for the accumulated experience of the participant’s close environment in daily undertakings. A study partner often provides information about the participant’s cognitive and functional performance; in a similar manner, we wanted to examine whether an informant’s HD observation could be a useful ‘marker’ of concomitant or future cognitive impairment. Study findings were consistent with our hypothesis. It is of note that findings do not support an increased risk for dementia in those with MCI and HD (compared with those with MCI but no-HD); informant-based HD reports might be a ‘marker’ for increased risk for someone to enter the cognitive impairment stage, but might not add additional information once someone is cognitively impaired and probably has more easily detectable clinical symptomatology. This is especially useful for early interventions, as informants’ observations could help assess individuals earlier in the dementia course.

The importance of current findings is underlined by the fact that there is an estimated 38 million persons in the US with hearing loss, which could reach 73 million by 2060 [25]; more than 40% of individuals older than 60 years are affected by hearing loss, and its significant clinical and socioeconomic consequences are underappreciated. Usually hearing assessment is not part of a routine primary care visit screening; often it is done once hearing loss has been noticed and hearing loss communication difficulties might not be thought of as a medical issue needing attention [1, 26]. However, if untreated, the cognitive, physical and psychosocial impact of hearing loss could be significant (e.g. impact on speech comprehension, reduced ability to communicate, reduced quality of life, isolation, loss of self-esteem, limitation in participation (in employment, in social events, in education), poor physical functioning and falls as well as, depression) [1, 16, 26]. Thus, the potential of hearing loss both to contribute and complicate dementia can be easily recognised [26].

In analysis by cognitive status (i.e. by CU and MCI) at baseline, the average yearly decline was modestly larger in those with MCI and HD. As suggested by others [16, 26], hearing loss could accelerate present subtle cognitive impairment as the cognitive burden increases and the relevant cognitive compensatory strategies might get exhausted. In addition, there could be volume decrease in the auditory cortex and hearing loss might accelerate brain atrophy (e.g. in the superior, middle and inferior temporal gyri and parahippocampal gyrus, areas commonly involved in Alzheimer’s); histopathological studies have also shown plaques and tangles in the auditory system of patients with AD [16].

Although the link between cognitive abilities and auditory acuity is not yet well understood, if hearing could be improved, the resulting reduction in cognitive load may improve cognitive performance (e.g. improve speech understanding and reduce the cognitive load of processing degraded sound), increase brain stimulation or improve social participation [26–28]. Such benefits would be very welcome considering that 9% of dementia cases could be attributed to midlife hearing loss [29]. Current treatments for hearing impairment (e.g. hearing aids, cochlear implants) can improve the auditory input, but remain underutilised, due to lack of knowledge of neglected hearing loss consequences or treatment choice options, high cost, technology challenges or incorrect use [16, 26, 27]; unfortunately, treatment delay makes it very difficult to regain previous hearing levels [1]. Whether early aggressive hearing loss treatment is effective for slowing or stopping cognitive decline remains to be answered [3, 26].

A limitation of the current study is whether informant reported HD could be due to cognitive difficulties or a combination of hearing and cognitive difficulties, as objective testing data (e.g. audiometric testing data) or use of hearing aids data were not available in MCSA. However, using the criteria (mentioned in methods) in neuropsychological testing to consider a participant as possibly having cognitive impairment, and adjudicating a cognitive diagnosis (CU, MCI or dementia) with consensus of three MCSA evaluators after review of the data (including hearing impairment if observed or reported) reduces the likelihood of a false diagnosis. If subtle cognitive changes yet undetectable by objective testing are present, the informant HD report that can trigger further hearing and cognitive testing is beneficial to the individual. On the other hand, researchers also report that overdiagnosis of cognitive decline might occur, if hearing impairment impacts performance on neuropsychological testing [30], in which case hearing acuity needs to be considered in cognitive testing.

The association of HD with increased risk of MCI and dementia is of particular interest and public health significance, as the burden of both hearing loss and dementia will increase with the aging population, with high personal, social and economic burden. Informant-based HD is easily obtained and could contribute in the early detection of HD, help in guiding the selection of persons for further evaluation—potentially those at risk for progression to MCI or dementia—and lead to interventions while individuals are still unimpaired cognitively or only mildly impaired.

Declaration of Sources of Funding

The study was supported by National Institutes of Health (NIH) Grants U01 AG006786, P50 AG016574, R01 AG041851, the GHR Foundation, the Alice Weiner Postdoctoral Research Fellowship in Alzheimer’s Disease Research, the Mayo Foundation for Medical Education and Research and was made possible by the Rochester Epidemiology Project (R01 AG034676).

Declaration of Conflict of Interest

Jeremiah Aakre, Razan Al Fakir, Chaitanya Undavalli, Rosebud Roberts: no conflict of interest. Maria Vassilaki receives research funding from NIH, Roche and Biogen. Walter Kremers receives research funding from the Department of Defense, NIH, Astra Zeneca, Biogen and Roche. Michelle Mielke consults for Eli Lilly, receives unrestricted research grants from Biogen, Lundbeck and Roche and research funding from the National Institute on Aging, NIH and the Department of Defense. Yonas Geda receives funding from the NIH and Roche and serves on the Lundbeck Advisory Board. Mary Machulda receives research funding from the NIH. David Knopman serves on a Data Safety Monitoring Board for the DIAN study; is an investigator in clinical trials sponsored by Lilly Pharmaceuticals, Biogen and the Alzheimer’s Treatment and Research Institute at USC and receives research support from the NIH. Ronald Petersen is a consultant for Roche, Inc., Biogen, Inc., Merck, Inc., Eli Lilly and Company and Genentech, Inc. and receives publishing royalties from MCI (Oxford University Press, 2003) and research support from the NIH.